AI in EBS: What's new and how it compares with Fusion

Key Concepts:

• EBS (E-Business Suite)
• Fusion
• AI (Artificial Intelligence)
• Generative AI
• Traditional AI
• OCI (Oracle Cloud Infrastructure)
• Select AI
• Apex
• ECC (Enterprise Command Centers)
• LLM (Large Language Model)
• Metadata
• Archiving Solution
• Unstructured Data

1. EBS and Oracle's Approach to AI

• Oracle is introducing AI to EBS, prompting a comparison with AI in Fusion.
• Inaps, a global Oracle systems integrator, has been exploring this since Cloud World in Vegas.
• Oracle delivered on their AI promises for EBS at the end of 2024.
• AI is pervasive, exemplified by using ChatGPT for job description creation.
• Oracle's AI strategy involves a stack with hardware, platform tools/services, and embedded AI in modern applications (Fusion, Analytics, NetSuite).
• EBS customers often have heavily customized systems built over 10-20+ years.
• While innovations like ECC are appreciated, they often address problems already solved by EBS customers through customization or third-party tools.
• There's excitement around the delivered AI in EBS, specifically generative AI.
• Select AI allows natural language queries against EBS data (e.g., "Tell me the suppliers who shipped late from Taiwan last week").

2. Traditional AI vs. Generative AI

• Traditional AI: Includes chatbots, voice recognition, document recognition, and RPA. Focuses on pattern analysis (e.g., identifying a tree from a picture).
• Generative AI: Creates something new from existing data (e.g., summarizing information, generating code). Can generate new pictures from existing images.
• Example: Using pictures of trees and flowers to generate a new picture.
• Concerns exist regarding copyright infringement when copyrighted material is used to train generative AI models.

3. How Select AI Works in EBS

• Delivered via a technical white paper (My Oracle Support note provided).
• Oracle provides an Apex page as the user interface.
• The process mirrors ECC installation, generating objects in an autonomous database.
• These objects are then written back into the EBS database, creating a new AI schema.
• Metadata about these objects resides in a new 23 AI database.
• When a natural language question is asked in Apex, it goes to the 23 AI database, calls the Select AI service, and generates SQL.
• This SQL is then run against the new AI schema in the EBS database.
• Analogy: Similar to creating views in a database to simplify reporting by combining data from multiple tables.

4. Oracle's Considerations for Implementing Select AI in EBS

• Requires EBS 12.2.7 or above.
• Requires access to OCI.
• Uses the Meta Llama large language model.
• The delivered solution does not include customizations.
• However, the process is based on ECC, and ECC can be extended to include customizations.

5. Demonstration of Select AI in EBS (Drew's Demo)

• Application allows questions by domain (e.g., accounts payable).
• Example: Asking for the total amount paid by each supplier returns supplier name and total invoice amount.
• Natural language is used to add criteria (e.g., "add currency").
• Drilling down into specific suppliers (e.g., "show all the invoices for the National Basketball Association").
• Asking for the top 10 highest values for a supplier.
• Asking "where are the suppliers located" correctly identifies the "supplier site" column.
• Asking "display more info" retrieves additional columns from the data set.
• The application shows the SQL query generated.

6. Comparison to AI in Fusion

• Oracle embeds AI in modern applications like Fusion.
• AI Assist: Summarizes data or creates goals.
• AI Agents: Run throughout a process (e.g., recruitment, procurement).
• Example: AI in Oracle HCM Cloud helps managers summarize peer feedback for performance reviews.
• Oracle states that the technology in EBS doesn't support the same level of AI embedding as Fusion.
• Fusion has guardrails for security, data leakage prevention, and compliance with government regulations.
• AI in Fusion is often included at no extra cost for SaaS customers.

7. Oracle's AI Strategy Summary

• OCI allows building custom AI.
• OCI provides pre-built AI services (e.g., Select AI).
• Oracle embeds AI in SaaS applications (Fusion).
• Embedded AI in Fusion is improved, secured, and supported by Oracle.

8. Select AI in EBS: Use Cases and Considerations

• Putting Select AI in EBS was straightforward.
• The initial release was US-centric.
• Not all modules are supported yet (e.g., procurement).
• Best use case: Speeding up development by allowing business users to have a conversation with their data and then provide the SQL to developers.
• Addresses the common problem of miscommunication between business users and developers.

9. Archiving Solution with Select AI (Drew's Demo)

• Inaps offers an archiving solution for EBS data after migrating to SaaS.
• Addresses the challenge of decommissioning old EBS systems.
• AI is enabled in the archiving tool using Oracle Select AI functionality.
• Example: Requesting payables invoices for a supplier over a date period.
• Retrieving all information on a specific invoice.
• Looking for a specific supplier's address details.
• Allows natural language querying of archived data without code.

10. Unstructured Data and AI (Drew's Demo)

• Demonstrates querying unstructured data (documents) using AI.
• Example: Summarizing a supplier's contract.
• Drilling down into specific contract clauses (e.g., payment terms).
• Finding other documents with similar information.
• Elaborating on specific sections of a document.
• The entire conversation takes place in Microsoft Teams, querying contract files in the Oracle database.

11. Conclusion

• The webcast focused on Select AI in EBS and its potential use cases.
• Compared AI capabilities in EBS to those in Fusion.
• Showcased the use of AI with structured and unstructured data.
• Highlighted the potential of AI to improve development processes and data access.
• Encouraged viewers to explore AI further and contact Inaps for assistance.